Organic Process Research & Development 2004, 8, 33−44
A New Approach to Rapid Parallel Development of Four Neurokinin Antagonists. Part 4. Synthesis of ZD2249 Methoxy Sulfoxide Sharon A. Bowden, J. Nigel Burke,† Fiona Gray, Steven McKown,† Jonathan D. Moseley,* William O. Moss, Paul M. Murray, Matthew J. Welham, and Maureen J. Young AstraZeneca, Process Research and DeVelopment, AVlon Works, SeVern Road, Hallen, Bristol, BS10 7ZE, UK, and Large Scale Laboratory, Macclesfield, UK
Abstract: The manufacture of ZD2249 methoxy sulfoxide (1) using a new project approach is described. Research department processes were scaled up to 100 L if process safety and robustness were not compromised; other factors were treated according to the new approach. Using this strategy, we were able to manufacture a key intermediate on sufficient scale to support delivery of 1 kg quantities of bulk drug within 6 months of the start of lab work.
Introduction In our previous paper,1 we presented a new approach by Zeneca Pharmaceuticals to the rapid parallel development of several candidate drugs. An overview of the strategy was given and illustrated by examples taken from across the neurokinin (NK) project.2 This was illustrated in more detail by the subsequent papers concerning ZD6021 cyano acid3 and the final assembly sequence.4 This contribution and the next5 describe our experiences of the manufacture of other key fragments of molecules required for the NK programme and further exemplify this new approach. In this case, the specific challenges of ZD2249 methoxy sulfoxide manufacture using the new project approach are discussed in detail. ZD2249 was the second molecule in the NK programme to enter development. The naphthalene cyano acid fragment was identical to that of ZD6021, and repeat manufacture of this portion was uneventful, as presented previously.3 The central (N-methylamine) portion was identical for all molecules in the series, and the final assembly sequence has also been presented in this journal recently.4 The one difference between ZD6021 and ZD2249 was in the 4-methoxy † Large Scale Laboratory, Macclesfield. (1) Moseley, J. D.; Moss, W. O. Org. Process Res. DeV. 2003, 7, 53-57. (2) Bernstein, P. R.; Aharony, D.; Albert, J. S.; Andisik, D.; Barthlow, H. G.; Bialecki, R.; Davenport, T.; Dedinas, R. F.; Dembofsky, B. T.; Koether, G.; Kosmider, B. J.; Kirkland, K.; Ohnmacht, C. J.; Potts, W.; Rumsey, W. L.; Shen, L.; Shenvi, A.; Sherwood, S.; Stollman, D.; Russell, K. Bioorg. Med. Chem. Lett. 2001, 11, 2769 and Albert, J. S.; Aharony, D.; Andisik, D.; Barthlow, H.; Bernstein, P. R.; Bialecki, R. A.; Dedinas, R.; Dembofsky, B. T.; Hill, D.; Kirkland, K.; Koether, G. M.; Kosmider, B. J.; Ohnmacht, C.; Palmer, W.; Potts, W.; Rumsey, W.; Shen, L.; Shenvi, A.; Sherwood, S.; Warwick, P. J.; Russell, K. J. Med. Chem. 2002, 45, 3972-3983. (3) Moseley, J. D.; Moss, W. O.; Welham, M. J.; Ancell, C. L.; Banister, J.; Bowden, S. A.; Norton, G.; Young, M. J. Org. Process Res. DeV. 2003, 7, 58-66. (4) Parker, J. S.; Bowden, S. A.; Firkin, C. R.; Moseley, J. D.; Murray, P. M.; Welham, M. J.; Wisedale, R.; Young, M. J.; Moss, W. O. Org. Process Res. DeV. 2003, 7, 67-73. (5) Parker, J. S.; Smith, N. A.; Welham, M. J.; Moss, W. O. Org. Process Res. DeV. 2004, 8, 45-50.
10.1021/op030039z CCC: $27.50 © 2004 American Chemical Society Published on Web 12/05/2003
substitution of the methoxy sulfoxide fragment (1). This was analogous to the Pip sulfoxide fragment (1′) of ZD7944, a molecule with some structural similarities to the NK series, which had previously been in development as an oral treatment for asthma.6 The incorporation of the 4-methoxy substituent required a different synthesis from that of 1′ for the initial stages. Rather than develop a new synthesis, we had sought to apply the Research-based route wherever possible if it did not compromise safety or robustness, as outlined in our project strategy.1 Our route to methoxy sulfoxide (1) was based on the Research route2 and is shown in Scheme 1. Modifications and manufacturing experiences are discussed below. Results and Discussion Bromophenol Stage (3). The process inherited from Research involved an apparently radical-mediated bromination of 3-methoxyphenol (2) in 1,1,1-trichloroethane, driven to completion by irradiation with a sun lamp that brought the mixture to reflux. After neutralization with sodium bicarbonate, the solvent was distilled off and the crude concentrate distilled under reduced pressure to yield the crude bromophenol (3) in 66% as a red oil. In spite of our stated aim of adopting Research-orientated processes for an expeditious manufacture, this stage had several undesirable features for scale-up, which we felt could easily be removed. The lack of control in a radically initiated bromination using benzoyl peroxide was a source of concern, as was the solvent, which is banned for use in manufacture under the Montreal Protocol.7 High-temperature vacuum distillation was also better avoided. Early sighting experiments showed that the process did not need the radical initiator or the sun lamp (6) Shenvi, A. B.; Jacobs, R. T.; Miller, S. C.; Ohnmacht, C. J.; Veale, C. A. U.S. Patent 5,589,489, 1996. Shenvi, A. B.; Aharony, D.; Brown, F. J.; Buckner, C. K.; Campbell, J. B.; Dedinas, R. F.; Gero, T. W.; Green, R. C.; Jacobs, R. T.; Kusner, E. J.; Miller, S. C.; Ohnmacht, C.; Palmer, W.; Smith, R.; Steelman, G.; Ulatowski, T.; Veale, C.; Walsh, S. Abstracts of Papers, Part 1, 214th National Meeting of the American Chemical Society, Las Vegas, NV, Sept. 7-1, 1997; American Chemical Society: Washington, DC, 1997; MEDI 264. Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
33
Scheme 1
and could in fact be run in the more acceptable dichloromethane at 20 °C with the exotherm controlled by the bromine addition rate. The distillation was avoided by successful telescope into the following crystalline O-thiocarbamate stage, as discussed below.
The main challenge of this bromination process was to control the ratio of the desired bromophenol (3) to the dibromide (12) (and other multiple bromination products), whilst achieving adequate conversion of 2. The Research conditions gave a 12:1 ratio in favour of the product with some residual 2 (which is an oil and is lost in the mother liquors at the next stage). Our change to dichloromethane gave a slightly worse but still acceptable ratio.8 We briefly investigated alternative conditions (standard electrophilic bromination,9 NBS in acetonitrile,10 and NBS in aqueous base11), but all gave much higher levels of regioisomeric diand tribrominated products as determined by LC-MS. To establish that we were working in a robust region of reaction space,1 and possibly to optimize the reaction further, particularly the ratio, we conducted a factorial experimental design (FED) using a four-cell HEL Automate reactor. We investigated five parameters at two levels in eight experiments, resulting in a quarter factorial. The parameters (7) Under Article 2E of the Montreal Protocol (United Nations Environment Programme, 1988), production and use of 1,1,1-trichloroethane, which is an ozone-depleting chemical, should have been phased out in developed countries by 1996. (8) We also tried chloroform which gave a marginally better ratio than that with dichloromethane, but not sufficiently so in our view to justify its usage. (9) Sandin, R. B.; McKee, R. A. In Organic Syntheses; Blatt, A. H., Ed.; John Wiley and Sons: New York, 1943; Collect. Vol. 2; pp 100-101. (10) Carreno, M. C.; Garcia Ruano, J. L.; Sanz, G.; Toledo, M. A.; Urbano, A. J. Org. Chem. 1995, 60, 5328. (11) Auerbach, J.; Weissman, S. A.; Blacklock, T. J.; Angeles, M. R.; Hoogsteen, K. Tetrahedron Lett. 1993, 34, 931. 34
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development
Table 1. FED parameters for bromophenol stage parameter
low level
high level
bromine charge (equiv) dichloromethane charge (volumes) addition temperature (°C) addition time (min) reaction temperature (°C)
0.75 6 5 15 5
0.95 8 20 60 20
investigated with their high and low levels are shown in Table 1; however, the fast addition rate and low-temperature combinations were deliberately excluded from the design as being unsuitable. The FED showed that longer addition times moderately favoured better selectivity, whilst other factors had little effect. Thus, we chose the lower temperature to control the exothermic addition, and 0.85 equiv of bromine as a compromise between unreacted 2 and dibromide (12). However, larger scale trials showed that the lower charge of bromine at 0.75 equiv with stirring at room temperature for several hours gave a moderately improved solution yield of 3. No other work specific to this stage was undertaken as it was successfully telescoped into the O-thiocarbamate stage. O-Thiocarbamate Stage (4). The Research process redissolved the distilled bromophenol (3) in DMF to which DABCO and dimethylthiocarbamoyl chloride (DMTCC) were added sequentially in portions, to acylate the free phenol. After an overnight hold, the crude O-thiocarbamate was precipitated by the addition of a large volume of water (25 volumes), and the crude dried solid then recrystallised from hot methanol in 44% overall yield for the two stages. We were concerned about the potential toxicity of DMTCC by comparison with the O-analogue, dimethyl carbamoyl chloride, a known animal carcinogen12 generated at low levels during the preparation of other NK compounds.1,4 The advice from our occupational health department was that the risk from the nonvolatile S-analogue was much lower and could be used under standard precautions for lab work and Large Scale Lab (LSL)13 manufacture. (12) Irving Sax, N. Dangerous Properties of Industrial Materials, 5th ed.; Van Nostrand Reinhold Company: New York, 1979.
We had few other concerns with this stage but felt that several worthwhile improvements could quickly be assessed. The first of these was to prove the reaction could be performed in dichloromethane, which allowed us to telescope from the previous stage, hence avoiding the distillation of the bromophenol. We cut down the equivalents of both DABCO (2.4 to 1.25) and DMTCC (2.4 to 1.03) without compromising the result. DMTCC also had much higher solubility in dichloromethane and so could be added as a solution, thus avoiding solid charging of a key reagent. A change to triethylamine from DABCO gave a gelatinous precipitate (presumably Et3N‚HCl) so that we retained the more efficient DABCO. An attempt to avoid the aqueous wash with dilute HCl to remove DABCO‚2HCl (now possible with dichloromethane in place of DMF) led to the precipitation of a white solid during subsequent processing and low-strength product, presumably both due to the presence of DABCO‚2HCl. The reaction was notably faster in DMF compared to dichloromethane (4 h versus 18 h at 20 °C), but the product form was poor and required drying before recrystallisation. Using NMP as a cosolvent (7% v/v with dichloromethane) did not improve the reaction rate and resulted in high losses to liquors. Finally, a solvent swap after the aqueous wash into methanol and distillation of the dichloromethane was successful, which avoided isolation and drying of the crude O-thiocarbamate. The product crystallized well from methanol, as noted by our Research colleagues, but was often slow to initiate. Seeding or self-seeding at ∼30 °C proved to be robust in the lab, and during manufacture only the first batch required seeding, the second two crystallising spontaneously. Three batches were manufactured in the LSL. Quality was typically >96% with the dibromo analogue (13) being the major impurity (untypically higher in the last batch). High product strengths indicated that UV-invisible inorganic and DABCO‚2HCl residues had been successfully removed. DSC analysis of the product 4 showed that there was no thermal instability at operational temperatures, and no further hazard testing was judged necessary in this case. The overall yield was ∼45%, effectively the same as the Research process but with simplified operations; about 15% product remained in the liquors which could not be isolated. Overall, we were able to apply standard process development improvements to these two stages and combine them into one telescope. Although the yield was moderate, we were able to produce adequate quantities of material in three batches. The processes had proved both safe and robust in manufacture and so had met the project remit, and further improvements were passed over in favour of working on later, more challenging stages as discussed below. (13) The Macclesfield LSL is a cGMP manufacturing facility for synthesis of bulk drug for clinical studies and uses all glass vessels. It is typically where the first significant scale-up of a process occurs, and commonly delivers tens of kilograms of intermediates and kilograms of bulk drug. It consists of a range of glass reactors 10-100 L in scale, fully contained with other ancillary equipment in fume cupboards. Operating ranges vary from -78 to +130 °C. Atmospheric hydrogenations can be performed, and a 20-L rotary evaporator is available for distillations if required. Product is generally isolated as a solid on Nutsches. AstraZeneca has several other LSLs at different sites which operate in a similar fashion.
S-Thiocarbamate Stage (5). This stage provided the first real test of applying the Research chemistry approach to this target molecule, as mentioned in our introductory paper.1 Formation of the S-thiocarbamate is achieved from the O-thiocarbamate precursor by a Newman-Kwart rearrangement14 using the method of Tı´ma´r15 in refluxing N,Ndiethylaniline (bp 218 °C). This could only be achieved in the LSL in a heating mantle, and that limited it to 20 L maximum. Previously for ZD6021 we had decided not to use a heating mantle for the cyano ester stage, although in part this was due to the presence of cyanide.3 That reaction had ideally required 180 °C, but only 130 °C could be achieved in a jacketed vessel, attenuating reaction times from 4 to 48 h. Necessity forced a rethink for the S-thiocarbamate process for which the higher temperature was essential. Fortunately, the process received from Research was also very concentrated, so that the manufacture could be comfortably completed in the 20-L flask in three batches, one of which was a smaller-scale proving batch. The Research process heated the O-thiocarbamate in degassed diethylaniline at reflux for 3-4 h; the solvent was distilled off by short-path distillation, and the residue was precipitated in ice-cold 6 M HCl, extracted with diethyl ether, dried, and concentrated to a brown solid which was recrystallised as a white solid from methanol in ∼60% overall yield. Aside from the operational use of the heating mantle which was new to the LSL at the time, we were concerned about the need for degassing the reaction solution, and its effectiveness in the LSL versus the lab. Lab preparations showed that degassing to remove oxygen was indeed necessary; without it, a lower yield of sticky-brown product higher in impurities resulted. Both evacuation/purging and sparging techniques were shown to be acceptable in the lab. In the LSL both methods were used in combination and were found to be successful without issue. If the impurity profile was good, it was found that a straightforward drown-out of the reaction mixture directly into an excess of ice-cold 6 M HCl could be used in place of the extractive procedure. This yielded the S-thiocarbamate as a light-brown solid which was washed with water and dried. Typical lab yields had been around 60% as for the Research process. In fact, the combined yield in the LSL for the 3 batches was 88%, well above the planning yield of 59%. Quality was also good, being typically 96% by LC area and >95% by strength. The major impurities were residual starting material (4), dibromoO-thiocarbamate (13), des-bromo-S-thiocarbamate (14), and dibromo-S-thiocarbamate (15), of which 4 and 15 were the two major ones (up to 3%, but only one predominated). Batch three had an untypically high level of 15 at 9.3%; all other impurities were below 1% by HPLC. Finally, preliminary hazard work of a DSC on the product 5 revealed a possible concern from 181 °C, well below the proposed reaction temperature. A more thorough, larger-scale Carius tube test revealed a prolonged double exotherm from 228 °C, quite possibly the heat of reaction since this is close to the temperature of the rearrangement. Although this was (14) Newman, S.; Karnes, H. A. J. Org. Chem. 1966, 31, 3980 and Kwart, H.; Evans, E. R. J. Org. Chem. 1966, 31, 410. (15) Sebo¨k, P.; Tı´ma´r, T.; Eszenyi, T.; Patonay, T. Synthesis 1994, 837. Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
35
judged acceptable on the 20-L scale, additional hazard testing would have been required for further scale-up. Fortunately, this did not present us with a capacity issue due to the concentrated nature of the process. Bromosulfide Stage (6). This stage was inherited as a partial telescope from Research and was already relatively simple. Hydrolysis of the S-thiocarbamate in methanol with 7.6 equiv of KOH gave the intermediate thiol (16) as an oil after an acidic workup, which was redissolved in DMF and treated with K2CO3 and methyl iodide to give the desired bromosulfide. The quantity of base seemed excessive, and this was readily reduced to 2.2 equiv, whilst the methyl iodide charge was cut down from 1.7 to 1.1 equiv (1.0 being just insufficient). The solvent swap to DMF also seemed unnecessary, so that the alkylation reaction was also conducted successfully in methanol. On completion, water was added and the product extracted into MTBE and concentrated to an oil. Hazard studies (Carius tube) showed that, although the hydrolysis phase was essentially an “all-in” process, there were no exotherms of concern; the alkylation reaction was effectively under process control by the methyl iodide addition.
This process was developed for manufacture in only 2 weeks, well inside our 4-6 week target. Other than the changes noted above, dimethyl sulfate was investigated as an alternative alkylating agent. However, methyl iodide gave a slightly faster, cleaner reaction so that this was retained, other safety and health issues being considered well controlled in the LSL environment. Quality was as good (>98% by HPLC), except for that of batch three where the high level of dibromo-(S)-thiocarbamate carried through as the dibromosulfide (17) at 6% by HPLC. Crude yield of the concentrated oil for each of the three batches was >100%, but with quality so high, we assumed this was residual MTBE. The product was an oil, and we had no initial standard with which to compare it. Unfortunately, the first batch of the following Grignard reaction was partially quenched, and we assumed from this that the residual solvent was, in fact, methanol. Azeotroping the bromosulfide batches from toluene under vacuum distillation reduced the yield to a more plausible 90% each and allowed the CBz alcohol stage to proceed smoothly. Our rapid development had been too rapid in this case. CBz Alcohol Stage (8). This stage proved to be the most demanding test of the new strategy on scale-up, since purification by chromatography was required. Initially, Grignard conditions were used in coupling bromosulfide (6) 36
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development
with CBz piperidone (7), by analogy with the nearly identical Grignard coupling of ZD7944 bromosulfide (6′) with CBz piperidone.6 Using 1.35 equiv of 6 in 1:1 toluene:THF gave a poor conversion of 6 to CBz alcohol (8) in only 30% yield after chromatography after several attempts. As noted in our introductory report,1 we used Kepner-Tregoe analysis for problem solving at critical points, and this proved to be effective here. Several potential causes for the low yield were proposed, but most were readily eliminated as being inconsistent with other experimental data. As discussed above, it was quickly shown by this methodical approach that residual solvent (methanol) in bromosulfide (6) was the cause of the problem (by partially quenching out the Grignard reagent). Azeotroping 6 with toluene removed the residual methanol, improving the yields significantly in the subsequent reaction. Whilst investigations had quickly been proving this hypothesis, we had also tried the harsher direct lithiation of bromosulfide with butyllithium at -78 °C, which appeared to work better. With the toluene-azeotroped bromosulfide, the lithiation reaction was quickly optimized in the lab. Neat anhydrous THF was found to be better than the tolueneTHF mixtures we had started with. Using a 1.2 equiv excess of 6, a 65-70% yield of 8 could be obtained after chromatography. These lithiation conditions were used for the first batch in the LSL, but unfortunately a troublesome unknown impurity was produced in this batch at 15% which had previously only been seen at